HStream

Easiest interactive python web app using htmx and semantic html

Usage

pip install hstream

# main.py

from hstream import hs

visitor_name = hs.text_input("What's your name?", default_value = "friend")

hs.markdown(f"Welcome {visitor_name}")

python -m hstream main.py

Motivation

Love Streamlit but:

find it hard to deploy
impossible to customise beyond PoC phase
overly complicated

H-(html)-Stream stream is built with semantic html, FastApi and htmx to provide a fast and simple framework for rapid web app development that follows traditional frontend/server architecture (or at least follow it closer than Streamlit).

Features

only reloads changed components after the visitor provides input
live server reload on file change (through univorn)
semantic html and basic html manipulation from within script
basic components - see below
swap stylesheet
auto ssl certs for easy deployment
complex html manipulation from within script (setting attributes)
plotly plot supprt

Bugs

key handling is a little inconsistent and it's not clear to the user when they need to use keys - meaning if you run into reloading / rending issues provide all you compoennts with unique key parameter
state management is a little wonky when scripts branch off a change in a component to create a new component
checkbox is a little wonky and doesn't retain state across refreshes (which happens when a component is added)

Components

hs.text_input

hs.checkbox

hs.slider: numeric slider input

hs.plot: output matplotlib figures to the user

hs.image: display an image

hs.html: allows more complex formatting, for example

hs.stylesheet_href = https://unpkg.com/@vladocar/basic.css@1.0.3/css/basic.css to use a different classless css framework

with hs.html('form'):
    hs.text_input('Name'):
    hs.checkbox('Would you like to be my friend?')

Technologies

Big thanks to the following libraries in particular

Streamlit
htmx
Yattag
MVP.css
FastAPI
uvicorn

richardsonjf / hstream